The Simple Way to Build Custom Actions for GPTs Without Coding (OpenAI Custom GPTs)

Liam Ottley
19 Mar 202412:18

Summary

TLDRThis video tutorial demonstrates how to create custom tools for GPT and AI agents without coding, using make.com and a webhook to connect with GPT. It showcases a step-by-step process to build an automation for sending emails through a GPT, utilizing a simple prompt template and OpenAPI schema. The tutorial emphasizes the potential of make.com for automating workflows and selling services, providing a valuable skill for both GPT users and those looking to monetize their automation expertise.

Takeaways

  • 🛠️ The tutorial demonstrates how to create custom tools for GPT and AI agents without coding, using make.com and a webhook.
  • 🔗 Make.com is highlighted as a favorite automation platform due to its user-friendly graphical editor and potential for complex automations.
  • 📧 The example given is about creating an email assistant using GPT and make.com to automate email sending.
  • 💰 Chat GPT Plus is required for this tutorial, which costs $20 per month.
  • 📝 The process involves defining a custom action ('send email operation') and prompting the user for recipient email and message content.
  • 🔄 A webhook is set up on make.com to interact with the GPT, and an OpenAPI schema is created for the custom action.
  • 📌 The tutorial references aGBT action schema Creator and a make.com OpenAPI template provided by Chris Lanon.
  • 🔗 The webhook URL and necessary inputs (recipient email, subject line, email body) are included in the schema.
  • 🚀 A test is conducted to 'fire' the webhook and determine the data structure, which helps make.com understand the expected inputs.
  • 🔄 Once the data structure is determined, the make.com scenario can use the information to perform the email sending action.
  • 🎯 The tutorial concludes with a live test, showing the successful integration and execution of the automated email sending process.

Q & A

  • What is the main topic of the video?

    -The main topic of the video is a tutorial on creating custom tools for GPT and AI agents without coding, using make.com and connecting them via a webhook to GBT.

  • Which platform does the video recommend for building custom automation tools?

    -The video recommends using make.com for building custom automation tools due to its user-friendly graphical editor and capability to create complex automations.

  • How does the video suggest integrating make.com with GPT?

    -The video suggests integrating make.com with GPT by using a custom webhook and creating an open API schema to connect the two platforms.

  • What is the example use case presented in the video?

    -The example use case presented is creating an email assistant that sends emails from the GPT using an automation set up on make.com.

  • What are the three fields required for the Gmail send action in the make.com scenario?

    -The three fields required are recipient email address, subject line, and email body.

  • How does the video address the challenge of authentication in creating open API schemas?

    -The video suggests using a webhook method that doesn't require authentication, making it easier for beginners to get started with building custom tools.

  • What is the purpose of the test fire in the make.com scenario?

    -The purpose of the test fire is to send real data to the webhook so that make.com knows what to expect and can use the data as variables within its automation.

  • How does the video help users debug their open API schemas?

    -The video provides a template for creating an open API schema and suggests filling in as much information as possible, especially the operation ID and parameters, to help with debugging.

  • What is the significance of the lightning bolt symbol in the make.com scenario?

    -The lightning bolt symbol indicates that the make.com scenario is awaiting new data, and it will trigger the automation process once new data is received.

  • What additional resources does the video recommend for learning about pre-built tools and custom actions?

    -The video recommends checking out a complete video on the Appify platform for learning about using pre-built tools and connecting them to GPT without the need for building custom tools.

  • Where can viewers find the GPT action schema creator and template mentioned in the video?

    -Viewers can find the GPT action schema creator and template on the video creator's free resource hub, which is accessible to members of their community.

Outlines

00:00

🛠️ Introduction to Custom Tool Creation with Make.com

The video begins with an introduction to creating custom tools for GPT and AI agents without coding, using Make.com and a simple prompt template to establish an open API schema for connection. The speaker emphasizes Make.com's ease of use and its potential for complex automations, suggesting its value for both personal and commercial applications.

05:01

📧 Setting Up an Email Assistant with Chat GPT

The speaker guides viewers through setting up a new GPT called 'email assistant' using Chat GPT's site, which requires a subscription. The tutorial focuses on creating an automation that sends emails via Make.com. The process involves defining a custom action for sending emails, prompting the user for recipient details, and determining the subject line autonomously.

10:02

🔗 Integrating with Make.com and Testing the Webhook

The tutorial continues with integrating the email assistant with Make.com by creating a custom webhook and adding a Gmail 'send email' module. The speaker explains the need to capture user input and fill in certain fields automatically. A test is conducted to determine the data structure and ensure the webhook's functionality. The speaker also acknowledges the contributions of Chris Lanon and provides resources for further learning.

Mindmap

Keywords

💡Custom Tools

Custom Tools refer to the unique applications or functionalities created for specific purposes, in this case, for GBTs (Generative Bot Tools) and AI agents. The video tutorial demonstrates how to create these tools without coding, using the platform make.com, which is highlighted for its user-friendly graphical editor that allows for complex automations over time.

💡GBTs (Generative Bot Tools)

GBTs, or Generative Bot Tools, are AI-powered applications that can generate responses or perform tasks autonomously. These tools are often used in customer service, content creation, and various other scenarios where automated responses or actions are beneficial. The video focuses on integrating these tools with make.com to enhance their capabilities without coding.

💡AI Agents

AI Agents are software programs or systems that are designed to perform specific tasks autonomously, often using artificial intelligence to make decisions or carry out actions. In the context of the video, AI agents are being integrated with custom tools created on make.com to automate workflows and potentially offer services to others.

💡Make.com

Make.com is an online automation platform that allows users to create complex workflows and automations without the need for coding. It provides a graphical editor for building automation scenarios, which can be connected to various tools and services, such as GBTs, to create custom functionalities.

💡Webhook

A Webhook is a way for an application to provide other applications with real-time information. It is a simple HTTP callback that allows for the transfer of data from one service to another when a specific event occurs. In the video, a webhook is used to connect a GBT with a make.com scenario, enabling the automation of sending emails.

💡Open API Schema

An Open API Schema is a standardized way of describing RESTful APIs that enables different software components to communicate with each other. It is used to define the structure of data and the operations to be performed, allowing for seamless integration between services. In the context of the video, the Open API Schema is crucial for defining how the GBT will interact with the make.com scenario.

💡Automation Workflows

Automation Workflows are a series of steps or actions that are automatically executed in a specific sequence to accomplish a task or a set of tasks. They are designed to streamline and optimize processes, reducing the need for manual intervention. The video tutorial focuses on creating such workflows on make.com that can be connected to GBTs and AI agents to perform various automated tasks.

💡Chat GBT Site

The Chat GBT Site is a platform where users can create and manage Generative Bot Tools. It is where the user in the video script goes to create a new GBT for the purpose of sending emails through an automation process. This site is part of the infrastructure that enables the creation and customization of AI-driven chatbots and other related tools.

💡Custom Action

A Custom Action in the context of the video refers to a specific function or operation that is not part of the standard set of actions provided by a platform. It is a user-defined action that extends the capabilities of the platform. For example, the 'send email operation' is a custom action that the user wants to implement for their GBT, allowing it to send emails through an automation on make.com.

💡Template

In the context of the video, a Template is a pre-defined structure or set of instructions that can be used as a starting point for creating something new. It simplifies the process by providing a framework that can be modified or customized to fit specific needs. The video mentions using a template for creating an Open API Schema, which is then customized for the email sending functionality.

💡Test Fire

Test Fire in the context of the video refers to the process of testing the functionality of the webhook by sending a real data request to it. This step is crucial for determining the data structure and ensuring that the webhook can correctly handle the inputs and execute the intended actions.

Highlights

The video provides a tutorial on creating custom tools for GPT and AI agents without coding, using make.com and a simple prompt template.

The method involves connecting to GPT via a whip hook and creating an open API schema for seamless integration.

Make.com is highlighted as a favorite automation platform due to its user-friendly graphical editor and potential for complex automation.

The tutorial example demonstrates how to send emails from GPT using an automation on make.com.

Creating a new GPT requires Chat GPT Plus subscription, which costs $20 per month.

The custom action for the email operation is 'send email operation', a technical term used by Open AI.

The tutorial emphasizes the importance of the web hook as the interaction point between GPT and make.com scenarios.

The Gmail send an email module is used to capture necessary information from the user, such as recipient email and message content.

A custom web hook is created and named, which will be used to interact with the make.com scenario.

The video credits Chris Lanon for the inspiration and resources used in creating the GPT action schema.

The open API schema is created using a template and information from the web hook to define the expected inputs and operation.

A test fire of the web hook from GPT is necessary to establish the data structure and placeholders for the automation.

The video provides a link to a free resource hub with additional templates and tools for further learning.

The process of determining the data structure within make.com is crucial for the automation to know what to expect from the GPT inputs.

Once the data structure is determined, the make.com scenario can be set to run on new data, automating the email sending process.

The video concludes with a live test of the automation, showcasing its successful integration and functionality.

The tutorial serves as an introduction to using make.com for beginners and provides a foundation for building more complex tools for GPT in the future.

Additional resources and a complete video on using pre-built tools with Appify are recommended for viewers interested in further learning.

Transcripts

play00:00

in this video I'll be giving you a super

play00:01

quick tutorial on how you can create

play00:03

custom tools for your gbts and for your

play00:05

AI agents without writing any code this

play00:08

method is using make.com and connecting

play00:10

them to your gbt via a whip hook and

play00:12

just using a super simple prompt

play00:14

template to create the open API schema

play00:16

that you need to connect them right so

play00:17

jumping into it the noode platform

play00:18

you're going to be building your tools

play00:20

on is of course make.com I've talked

play00:22

about this a lot on the channel before

play00:23

it is my personal favorite automation

play00:25

platform and I think it's got a a really

play00:27

good graphical editor so that you can

play00:28

drag it around very and friendly but

play00:30

also allows you to get really good at it

play00:32

over time and build some very very

play00:34

complex things so the skill you're about

play00:35

to learn now is going to be not only

play00:37

valuable for you gbts but also in how

play00:39

you can use make.com to automate

play00:40

workflows and sell their services to

play00:42

other people as well so make.com be a

play00:44

linked down below if you you can't type

play00:45

in make.com if that's a struggle um but

play00:47

you want to create an account and I'm

play00:49

going to log into mine here and then

play00:50

what we want to do is create a new

play00:52

scenario up here now that we have our

play00:53

new scenario we can head over and go to

play00:56

the chat gbt site and we can create a

play00:58

new GPT and sort of lay the for what

play01:00

we're going to connect to this so you

play01:01

get a little bit more context on what

play01:02

we're doing here so explore gpts you'll

play01:05

need chat GPT plus to do this $20 a

play01:07

month we're going to create a new gbt

play01:09

I'm going to call this an email

play01:11

assistant so the example I'm going to

play01:13

use in this tutorial is allowing us to

play01:15

send emails from our gbt using an

play01:18

automation that we put on make so first

play01:19

we'll go in here um sends emails for us

play01:22

via make.com custom action um when the

play01:26

user asks you to send an email the the S

play01:31

email operation so that's the custom

play01:32

action we're referring to that we're

play01:33

going to put in in

play01:37

second that's a technical name that open

play01:40

AI uses for a tool or custom action so

play01:42

use to send email operation um and ask

play01:46

the user for the email of the recipient

play01:50

and the email message then figure out

play01:53

the subject line yourself so that should

play01:56

be it we want this conversation started

play01:58

to be send email

play02:00

I let send an email so now that we've

play02:04

got the bones laid there you know what

play02:06

we're going to be building we can head

play02:07

over to Let's like get a little bit of a

play02:09

an image going here I always feel like

play02:11

it isn't complete until you put an image

play02:12

on it so now we can H back to make and

play02:14

we can actually lay this thing out so

play02:16

first thing is going to be a web hook

play02:18

this is going to be how our uh GPT can

play02:21

actually interact with our make.com

play02:23

scenario scenario is what this whole

play02:25

kind of like automation is called and

play02:27

make so we want to click a custom web

play02:29

hook here

play02:30

and we want to add a new web hook I'm

play02:32

going to call this U make

play02:36

tutorial or GPT you can call it whatever

play02:39

you want probably something to do with

play02:40

the gbt that you're creating and then

play02:42

we're going to go

play02:44

okay and then we want to add another

play02:45

module in this case as I said we're

play02:47

going to be doing a send an email

play02:50

capability so we want to go Gmail send

play02:53

an email you can see that we have three

play02:55

Fields here this so this is the

play02:56

information that we need to capture from

play02:58

our user or fill in some of of them

play03:00

ourselves as you saw in the prompt I

play03:01

said for the subject line don't expect

play03:03

the user to provide that we can figure

play03:04

it out I want the gbt to figure it out

play03:07

from the content and the message we need

play03:09

to get the email and the content from

play03:10

the user and the gbt is going to figure

play03:12

out the subject so in order to get the

play03:13

information in here as you can see here

play03:15

we need to be able to put some kind of

play03:16

variable in here right we need to have

play03:18

some some placeholders that have taken

play03:20

it from the we hook and place it in

play03:21

there but I won't be able to get that

play03:23

information until we've done a test fire

play03:25

we need to test fire this web Hook from

play03:27

our GPT here there we go nice picture so

play03:29

in order to do that we need to create

play03:31

our open API schema and add it in as a

play03:33

custom action down here I want to give a

play03:35

quick shout out to Chris lanon here it

play03:37

was actually this video that I saw that

play03:38

gave me the idea for this video so thank

play03:40

you Chris you beat me to it by about 3

play03:42

months I think and in this video Chris

play03:43

shared a gbt action schema Creator which

play03:45

we're going to use in this video which I

play03:47

found is quite helpful and there's also

play03:49

a template for a make.com open API

play03:51

scheme that we're going to be using as

play03:52

well so Chris thank you for the work

play03:54

that you put into this uh I appreciate

play03:56

and my community does as well so in here

play03:58

we can say create me a

play04:02

schema it's going to ask me for a bunch

play04:04

of information and we can sort of

play04:06

shortcut it here I say here is the

play04:10

template I'm

play04:12

using and now going back to what I

play04:15

mentioned we want to include in the

play04:16

schema that we need these three Fields

play04:18

so two subject and content if I go back

play04:21

one I want to copy this while we're at

play04:23

it copy this we want to bring this over

play04:26

uh we can say here is my web hook

play04:31

URL here are the inputs this uh tool

play04:37

requires and then you put in the inputs

play04:39

of

play04:42

recipient email

play04:46

address uh

play04:49

subject

play04:50

line email body so those are the three

play04:53

things that the Gmail is requiring and

play04:55

this is the webbook URL and what it's

play04:57

basically going to do is look through uh

play04:59

this template we've got and replace

play05:01

things like okay what's the summary of

play05:02

the action okay what's parameter one in

play05:04

this case it'll be the recipient email

play05:06

address re parameter 2 is going to be

play05:08

for the subject line and parameter 3 is

play05:10

going to be for the Emil body so if we

play05:12

enter this let's go and modify that to

play05:14

our needs and you can take the same

play05:16

process a link to Chris's GPT here and

play05:18

also the template included will be

play05:20

available at my free resource Hub on

play05:21

school so if you're not already a member

play05:23

of my school it's a free community week

play05:24

and get all the resources that are

play05:26

attached to my videos more importantly

play05:27

engage in the discussion and be in the

play05:29

loop on all the AI solutions that my

play05:31

community is selling successfully to

play05:32

business owners really just get closer

play05:34

to the action in terms of the AI space

play05:35

of people who are try to build

play05:37

businesses and make money so if you're

play05:38

not already a member of the school

play05:39

Community it's down there it's free um

play05:41

and on there if you go to the YouTube

play05:42

tab there'll be a post for this video

play05:44

and in it will be a link to the GPT

play05:46

action schema Creator here and also the

play05:48

template so you can continue to reuse

play05:49

that use the same kind of structure that

play05:51

I just provided where you can paste the

play05:52

template in you provide your make.com

play05:54

webblog URL and then you provide the

play05:56

inputs that are necessary and then it

play05:58

will generate you the schema for it oh

play05:59

it's giving me a bit of back chat here

play06:01

what I like to do is add to the bottom

play06:03

of this um fill

play06:07

in and the rest of the Fe

play06:12

field

play06:14

yourself as best as possible and modify

play06:17

this template to use the URL and inputs

play06:22

I've Pro provided I didn't really tell

play06:25

her what to do there so there we have

play06:26

our webook URL being filed operation ID

play06:28

is called send email good first

play06:31

parameter is recipient email yep the

play06:33

second parameter is subject line and the

play06:35

third one is email body might help you

play06:37

debug these in future if you know how to

play06:38

read them a little bit so we have the

play06:40

name of the operation um and the

play06:42

parameters that we expect as inputs now

play06:44

we can take this copy it head back to

play06:47

rgbt paste it in and because we're using

play06:49

a make webook we don't need to do any

play06:51

authentication so this is why I

play06:52

particularly like this method for

play06:54

beginners cuz authentication can get a

play06:55

little bit tricky in writing these

play06:56

schemers as well so what we can do now

play06:58

is hit back I like to give it a refresh

play07:00

just to make sure it's properly set up

play07:03

and then we can click on send an email

play07:05

here and what we need to do is determine

play07:08

the data structure within make we

play07:09

haven't filled in these variables yet

play07:10

with placeholders that's because we need

play07:12

to call this first and we need to send

play07:14

some real data to this webook so that it

play07:16

knows what to expect and we can use them

play07:17

as variables with then make sounds

play07:19

complicated but I promise it's not uh

play07:21

all we need to do here is say okay set

play07:23

the two field it's trying to call the

play07:25

toia it's asking for the email address

play07:27

and the message we like to send the two

play07:29

Mike at

play07:31

gmail.com

play07:33

message um you the bomb man I miss you

play07:39

so much why you know before I hit enter

play07:42

on that I want to go back to the webook

play07:44

I want to click redetermined data

play07:46

structure make sure that's make sure

play07:48

that's underway and then I send this

play07:50

it's going to ask me to approve the

play07:52

action yep so it's starting to try and

play07:55

send and communicate with our web hook

play07:56

then I want to we can see it's trying to

play07:59

send to mik Gmail subject line missing

play08:01

you which it's figured out which is

play08:02

great and the email body so if I

play08:05

allow and I go back this should change

play08:08

to successfully determine data structure

play08:10

which is exactly what we want give it a

play08:13

second boom successfully determined and

play08:16

that means that make has gone okay this

play08:18

structure that you sent I now know what

play08:20

to expect and I can give you these to

play08:21

use within make so this is really the

play08:23

key step and it can be difficult for

play08:24

beginers sometimes but I hope that it's

play08:26

been simple enough I've had to refilm

play08:28

this once just to make sure I do a

play08:29

little bit clearer so the email has been

play08:31

sent we're all good back over here now

play08:33

we have the information that's arrived

play08:35

within make I'm going to save it

play08:37

sometimes I need to play around with

play08:38

this a little bit but if you refresh

play08:40

it it should have loaded in and we can

play08:42

use it within our Gmail and then we can

play08:44

give this one final test and we're all

play08:46

done sometimes you have a bit of issue

play08:47

with the variables popping up like this

play08:49

so these aren't popping up for you um

play08:51

easy fix that I found is if you can

play08:53

unlink these unlink the two and then you

play08:55

just click on the Run once it's going to

play08:57

be waiting waiting waiting and I can

play08:59

just say say uh let's send email again

play09:03

and it's going to send some live data

play09:05

again give it a second and there should

play09:08

be a little tick in a

play09:10

second and this has come through uh we

play09:13

can see that we've got the data that's

play09:14

come through all good here and now when

play09:17

we connect these up I click on this and

play09:19

go there we go subject line we can fill

play09:21

that in content is email body add a

play09:24

recipient recipient email and now we

play09:27

have mapped the variables from the input

play09:29

that out tool is expecting from the

play09:30

inputs that it's taking from the gbt and

play09:33

it successfully mapped it in that it can

play09:34

easily slide those variables from the

play09:36

webook directly into the Gmail so once

play09:38

that's all set up I'm going to turn this

play09:41

scheduling on me go back a step save the

play09:44

changes so now with the scheduling on

play09:46

this means this is awaiting any new data

play09:48

with anything in make that has this

play09:49

lightning bolt it fires as soon as new

play09:51

data comes in so as soon as that we Pok

play09:53

is fired by a GPT it's going to run that

play09:55

process again so now I can get my phone

play09:57

out and I can give this a proper test if

play09:59

I go save I'm going to make this a

play10:02

little private

play10:04

one I'm going to view it on the big

play10:06

screen and I go send an email I'm going

play10:09

to get my phone out see if I can get an

play10:11

email from myself I go to my Gmail app

play10:15

here and if I

play10:18

go

play10:20

to send an email

play10:26

to and say hey hey man what H happened

play10:32

to

play10:34

our

play10:35

meeting did you change the time from

play10:40

Greg okay so we give that a second oh

play10:43

and if we go back to here we should see

play10:45

that it pinks it fires we're going to

play10:47

always allow

play10:51

it talk to the hook boom boom so these

play10:55

little bubbles mean that it's done

play10:56

something so receed the information here

play10:59

and then it sent the email off here

play11:03

from Etc sent it off and up here if I

play11:08

refresh meeting time inquiry and I've

play11:10

got my email there as expected so that's

play11:13

how you can integrate a make.com

play11:15

scenario as a tool this is something I

play11:17

wanted to introduce to you Bo because I

play11:19

know many of you use make from watching

play11:20

my videos makes a great thing you can do

play11:22

some really complex automations on here

play11:25

um so once you built up your skills on M

play11:26

you can transfer those over to creating

play11:28

custom tools for your GP

play11:29

and as someone who doesn't know how to

play11:31

code or you're not really too familiar

play11:33

with building your own custom tools this

play11:35

can be a great way to not only build

play11:36

them and a fairly simple system for

play11:38

connecting them via a schema as it

play11:40

doesn't require any fancy authentication

play11:41

so this system that I just showed you

play11:43

here of providing the inputs providing

play11:45

the webook URL asking it to modify the

play11:48

template again all of this will be on my

play11:49

free resource Hub and thank you to Chris

play11:51

for the information that he shared and

play11:53

this GPT in particular so that's about

play11:54

all for the video let me know down below

play11:56

if you found this helpful I definitely

play11:57

did and I'll be using this in some my

play11:59

future videos but aside from that if

play12:01

you're interested in learning the

play12:02

easiest way that I've found to use

play12:03

pre-made tools and connect them to your

play12:05

gbt so that you don't even need to do

play12:06

the building step I've just done a

play12:08

complete video on the appify platform

play12:09

here which is great for using pre-built

play12:11

tools using other people's hard work so

play12:13

highly recommend you watch that aside

play12:15

from that guys thank you so much for

play12:16

watching and I'll see you in the next

play12:17

one

Rate This

5.0 / 5 (0 votes)

Related Tags
Automationmake.comGPTAI AgentsWebhooksNo CodeWorkflowBusiness ToolsOpen APIEmail Automation